The Molecular Modeling Section develops and uses theoretical tools to study and predict the structure and stability of globular protein molecules and to design protein molecules with new or improved properties. It also analyzes the Expressed Sequence Tag (EST) cDNA database to discover new genes for use as targets for immunotoxins in cancer therapy. Following are some of the notable achievements made during this reporting period: (a) We began to develop a new method for recognizing protein structural types from its sequence alone (fold recognition). The method uses the pair-to-pair scoring matrix we developed in the previous year, as well as a new contact potential and a novel sequence-structure alignment procedure. The method is currently under active development. (b) The ab initio protein structure search engine that was partially developed in the previous year has been essentially completed this year by adding efficient local moves. We are now developing the energy function that will drive this engine. (c) A procedure was developed for systematically searching for circularly permuted protein structures. Using this procedure, it was found that nearly half of all known protein structural domains show evidence of having been circularly permuted with respect to some other protein in the database. This finding suggests that many of the protein domains have been produced by exon shuffling and gene duplication/fusion events. (d) We have our own in-house molecular graphics program, which is being heavily used. However this program was developed about 15 years ago and does not take advantage of the modern features of the graphics software and hardware. Instead of modifying the existing program, we recently decided to write a new molecular graphics program from scratch, using an object oriented programming language (C++) and modern software development tools. We have been able to make a good progress so far and plan to continue the development. (e) In cooperation with the Molecular Biology Section, we designed mutants of several different antibody Fv molecules for use as the cancer cell-targeting moiety of immunotoxins. These have lower toxic side effect, better yield, and/or more stability. (f) We devised databases and tools to classify and sort NCBI's Unigene clusters according to organ or cell-type specificity. This procedure now supersedes our own EST clustering procedure and saves a huge amount of computation time for us. Several promising new prostate-specific genes have been found using new lists generated by this procedure. (f) Our own EST clustering procedure has been automated and applied to the Unigene clusters to sub-cluster them. We found many alternatively spliced examples using this procedure, some of which turned out to be organ specific.